UAV_Navigation_DRL_AirSim
UAV_Navigation_DRL_AirSim copied to clipboard
heleidsn
This is a new repo used for training UAV navigation (local path planning) policy using DRL methods.
各位大佬，第一次接触airsim仿真，按照要求装了所需的环境之后，我在ubuntu下，下载了building_99数据集，双击.sh文件运行是可以运行的。但是运行训练文件python scripts/start_train_with_plot.py是报错，错误如下：File "scripts/start_train_with_plot.py", line 35, in main training_thread = TrainingThread(config_file) File "/home/xzc/UAV/UAV_Navigation_DRL_AirSim/scripts/utils/thread_train.py", line 54, in __init__ self.env = gym.make('airsim-env-v0') File "/home/xzc/miniconda3/envs/airsim/lib/python3.8/site-packages/gym/envs/registration.py", line 669, in make env = PassiveEnvChecker(env) File "/home/xzc/miniconda3/envs/airsim/lib/python3.8/site-packages/gym/wrappers/env_checker.py",...
Hello, we have recently been training a drone obstacle avoidance model, and I am using the "simple avoid" scenario. However, the training results are not very good with the default...
我在使用深度强化学习训练多无人机避障时，将 ClockSpeed 设置为 5 来加快训练。当我完成训练后，如果我仍在 ClockSpeed 设置为 5 的环境下进行测试，无人机可以成功避碰。 但是当我将 ClockSpeed 设置为 1 进行测试时，无人机就会发生碰撞。我想请问一下这是为什么啊？难道ClockSpeed 设置会影响无人机本身的动力学吗？为什么在测试时也必须要设置跟训练时一样的ClockSpeed 值才能成功复现训练结果啊？如果是这样的话，那把策略迁移到实际环境中时不会出问题吗？毕竟真实环境下，应该相当于ClockSpeed=1的情况吧？希望大佬帮忙解答一下，非常感谢！
在运行gym.make('airsim-env-v0')时一直找不到action_Space(), ![Uploading 屏幕截图 2024-03-23 013225.png…]()
在训练时可以正常运行，在以logs目录下保存的文件夹为eval path运行evaluate的时候就会报错，报错如下： D:\anaconda\envs\uavsim\python.exe scripts\start_evaluate_with_plot.py Traceback (most recent call last): File "scripts\start_evaluate_with_plot.py", line 62, in main() File "scripts\start_evaluate_with_plot.py", line 36, in main gui = TrainingUi(config=config_file) File "E:\DRL\UAV_Navigation_DRL_AirSim-main\UAV_Navigation_DRL_AirSim-main\scripts\utils\ui_train.py", line 37, in __init__...
 I realized that the UAV lands randomly in the middle of the current iteration and it starts from there again in the next iteration. Is there a configuration time...
我想请问下，您在训练时，是否使用了Domain randomization的方法，来使得训练获得的策略具有更好的泛化性呢？比如每回合训练开始时，UAV随机初始化在不同的位置？
你好，我运行了AirSimNH.exe/SimpleAvoid.exe，之后再运行start_train_with_plot.py，但是在这两个环境下，都报错“No registered env with id”，请问如何解决这个问题？ 报错具体如下： (airuav) E:\UAV_Navigation_DRL_AirSim>python start_train_with_plot.py init training thread Traceback (most recent call last): File "E:\anaconda3\envs\airuav\lib\site-packages\gym\envs\registration.py", line 150, in spec return self.env_specs[id] KeyError: 'airsim-env-v0' During handling of...
`"D:\Program Installation\Anaconda\envs\Airsim_DRL\python.exe" D:\PycharmProjects\Airsim_DRL\scripts\start_train_with_plot.py init training thread Traceback (most recent call last): File "D:\PycharmProjects\Airsim_DRL\scripts\start_train_with_plot.py", line 53, in main() File "D:\PycharmProjects\Airsim_DRL\scripts\start_train_with_plot.py", line 35, in main training_thread = TrainingThread(config_file) File "D:\PycharmProjects\Airsim_DRL\scripts\utils\thread_train.py", line 55,...
环境切换
当我把简单环境切换为NH时出现了下面的错误  config为：config_NH_center_SimpleMultirotor_3D.ini 期待您的解答
